Probing the Quantum States of a Single Atom Transistor at Microwave

Nov 15, 2016 - The ability to apply gigahertz frequencies to control the quantum state of a single P atom is an essential requirement for the fast gat...
0 downloads 10 Views 4MB Size
This is an open access article published under an ACS AuthorChoice License, which permits copying and redistribution of the article or any adaptations for non-commercial purposes.

Probing the Quantum States of a Single Atom Transistor at Microwave Frequencies Giuseppe Carlo Tettamanzi,* Samuel James Hile, Matthew Gregory House, Martin Fuechsle, Sven Rogge, and Michelle Y. Simmons School of Physics and Centre of Excellence for Quantum Computation and Communication Technology, UNSW Australia, Sydney, New South Wales 2052, Australia ABSTRACT: The ability to apply gigahertz frequencies to control the quantum state of a single P atom is an essential requirement for the fast gate pulsing needed for qubit control in donor-based silicon quantum computation. Here, we demonstrate this with nanosecond accuracy in an all epitaxial single atom transistor by applying excitation signals at frequencies up to ≈13 GHz to heavily phosphorus-doped silicon leads. These measurements allow the differentiation between the excited states of the single atom and the density of states in the one-dimensional leads. Our pulse spectroscopy experiments confirm the presence of an excited state at an energy ≈9 meV, consistent with the first excited state of a single P donor in silicon. The relaxation rate of this first excited state to the ground state is estimated to be larger than 2.5 GHz, consistent with theoretical predictions. These results represent a systematic investigation of how an atomically precise single atom transistor device behaves under radio frequency excitations. KEYWORDS: silicon, single atom transistor, phosphorus, monolayer-doped electrodes, pulse spectroscopy, relaxation rates

A

scales. Control signals in the gigahertz regime are desirable for dispersive readout15 and for controlling exchange interactions for nonadiabatic gate operations.16 Indeed, a recently proposed scheme for implementing the surface-code error correction protocol in silicon relies on the ability to propagate signals through such devices with sub-nanosecond timing precision.17 Recent impurity-based quantum charge pump devices have been shown to be robust in terms of immunity to pumping errors when operated at gigahertz frequencies.18,19 However, to date, these experiments have been performed on devices containing random ion-implanted impurities.18,19 STM fabrication capabilities can allow high-precision (≲ nanometer) positioning of the dopant,1 and when combined with highspeed control of quantum states, it will provide devices for quantum metrology.18,19 In this paper, we investigate the propagation of highfrequency signals to the monolayer-doped leads used in atomically precise devices. Previous results have demonstrated the ability to apply radio frequency (rf) (≈300 MHz) transmission using dispersive measurements for manipulation of the quantum states.15 Here, we present a systematic study of the propagation of high-frequency signals in atomically precise

dvances in Si device fabrication technology over the past decade have driven the scale of transistors down to the atomic level. The ultimate limit of this scaling is to fabricate a transistor with just one single dopant atom as the active component of the device, and this has been realized using scanning tunneling microscope (STM) lithography.1 The spin states of individual P donor electrons and nuclei have extremely long coherence times when incorporated into a crystal composed of isotopically purified 28Si,2−4 making them excellent candidates for quantum information processing applications.5−7 STM lithography offers the potential to scale up such qubits by providing a means to position individual P atoms in a Si lattice and align them with sub-nanometer precision to monolayer-doped control electrodes. This technique has already demonstrated double8 and triple9 quantum dot devices, controllable exchange interactions between electrons,10 and the ability to initialize and read out the spin states of single electrons bound to the donor with extremely high fidelity.11 Most recently, these monolayer-doped gates were shown to be immune to background charge fluctuations, making them excellent interconnects for siliconbased quantum computers.12−14 Besides the ability to create devices with atomic precision, another requirement for quantum information processing and high-speed logic applications is the ability to control the quantum states of the donor electrons at sub-nanosecond time © 2016 American Chemical Society

Received: September 20, 2016 Accepted: November 15, 2016 Published: November 15, 2016 2444

DOI: 10.1021/acsnano.6b06362 ACS Nano 2017, 11, 2444−2451

Article

www.acsnano.org

Article

ACS Nano

Figure 1. High-frequency measurements of a single atom transistor. (a) STM image of the device. (b) Schematic of the measurement circuit showing how source−drain leads and two gates lines (G1/G2) are used to control the chemical potential of the donor, where a rf signal Vrf1/ Vrf2 is added to the conventional dc signal via bias tees. (c) Excitation spectrum of the D0 state of a P donor in Si (in color).20−22 The bulk values for the energy differences between these excited states (ESs) and the ground state (GS, blue) are also shown for the first ES (red), the second ES (green), the third ES (dark blue), and the fourth ES (orange).20−22 In this picture, the lowest three states (1s(A1); 1s(T2); 1s(E)) come from the linear combination of the Si valleys due to breaking of valley degeneracy in the Si lattice, and the last two (2p0 and 2p±) are orbital-like.22

Figure 2. High-frequency control of the D0 ground state using the gates (G1, G2). Square root of the power dependence of the position of the D+ to D0 current peak over ≳4 orders of magnitude change in frequency of an rf sine applied to G1 (VG1‑rf) from (a) 1 MHz to (c) 12.785 GHz, where T = 1.2 K, VSD = 2.6 mV, VGS = VG2 = VG1 + 400 mV and VGS0 is the position of the GS peak in dc (VG2‑0 ≈ 620 mV and VG1‑0 ≈ 220 mV). The signal becomes asymmetric above the 1 GHz frequencies, due to frequency-dependent cross-coupling between the gates and the source/drain leads giving rise to rectification effects. (d) Schematic describing the doubling of the D0 current peak is shown. As described in the main text, the green/red regions illustrate the positions available for the state when both rf and dc signals are in use.

first excited state to the ground state, ΓES, in good agreement with previous experiments25,26 and theoretical estimations.27 It is important to note that such a large range in the extracted value of ΓES can be linked to the strong tunnel coupling of the state to source/drain leads in this particular device, making the experiments needed for a more quantitative result infeasible.23,24 However, in the long term, this coupling can be controlled by the geometry of the tunnel junctions, which can

devices. In this work, we demonstrate high-frequency capacitive coupling (up to ∼13 GHz) to the states of a single atom transistor20,21 fabricated via scanning tunneling microscope lithography (see Figure 1a), important for the implementation of quantum information processing2−7,17 and quantum metrology.18,19 We report transient spectroscopy experiments23,24 that confirm the existence of the excited state of the P donor located at an energy of 9 ± 1 meV, and we extract bounds from 2.5 to 162 GHz for the relaxation rates from the 2445

DOI: 10.1021/acsnano.6b06362 ACS Nano 2017, 11, 2444−2451

Article

ACS Nano

GHz). In Figure 2d, we show a schematic describing how the doubling appears at different power and the underlying mechanisms causing it. When the rf signal is applied to one of the two gates (G1 or G2), during each rf cycle, the GS can occupy a range of positions represented by the green/red regions in the schematic of Figure 2d, where the green and red regions simply refer to the voltage change rate at which the donor GS crosses the bias window and depends on the timing of the sine wave (green = low rate of change of the sine; red = high rate of change of the sine). To clarify, at any point in time of the sine period, the current is proportional to the portion of integrated time that the states spend within the bias window. Hence, if the variation in time of

be engineered with sub-nanometer precision during fabrication.28

RESULTS AND DISCUSSION In contrast to surface gate-defined quantum dot devices, which typically make use of macroscopic metal electrodes to propagate high-frequency signals, atomic precision devices rely on electrodes formed using highly phosphorus-doped silicon (∼2.5 × 1014 cm−2), where the phosphorus dopants form a monatomic layer within the Si crystal patterned in the same lithographic step as the single donor atom (see Figure 1a). Within the monolayer of dopants, the average separation of the donors is ≲1 nm, giving rise to a highly disordered twodimensional electron gas. Disorder scattering in these degenerately doped leads gives rise to a resistance of hundreds of ohms per square, comparable to that found in silicon quantum dots29 but one order of magniture higher than the values observed in conventional transistors.30 However, another very important difference is that the self-capacitance of the atomically thin monolayer wires are negligible with the cross capacitances to the other leads being quite small, estimated to be around the aF.21 As a consequence, very little current (≈nanoampere) is required to carry a high-frequency voltage signal along these wires if compared to the tens of nanoamperes necessary for quantum dots.29 Figure 1 shows in (a) an STM image of the device and (b) a schematic of the measurement circuit used, illustrating how both dc and rf signals can be applied to gate 1 (G1) and to gate 2 (G2) via bias tees. The pink areas in Figure 1a show the highly P-doped monolayer regions (also see the Methods section) comprising tunnel coupled source/drain (S/D) leads and capacitively coupled gates (G1/G2) surrounding a single phosphorus atom. Several step edges separating the individual atomic planes are clearly visible in the STM image. To test the frequency response of the monolayer-doped gates, the D+ to D0 current peak related to current flow through the isolated P atom18,20,21 can be capacitively addressed by two gates (i.e., G1 and G2), allowing an independent rf signal to be added to each of the two gates and the device to be studied in both the dc and the rf domains. The use of rf signals is particularly attractive for these atomic-scale devices as the very narrow leads (≲5 nm) used to address the donor are quasi-1D, making it difficult by using simple dc bias spectroscopy to distinguish the signatures in current related to the excited states of the donor from the features related to the density of the states (DOS).20,21,31,32 Later, we will show how we apply transient current spectroscopy as described previously23,24 to clarify some of the transport mechanisms that can arise throughout the excited state spectrum of a single atom transistor. In Figure 1c, a schematic of the excitation spectrum of the D0 state of a P donor in Si22 is shown highlighting the 1s(A1), 1s(T2), and 1s(E) valley states and the 2p0 and 2p± orbital states of the single donor. In Figure 2, we observe the evolution of the current peak related to the ground state (GS) of the D0 state as a function of the power of the sinusoidal rf signal added to the dc voltage of gate 1. The possibility of capacitively addressing this D0 GS is confirmed for high frequencies up to ≈13 GHz, where, as expected, when an rf signal with sufficient power is in use, the position of the D0 current peak splits in two, with the splitting being proportional to the square root of the power of the provided excitation. This doubling of the current peak is observed for more than 4 orders of magnitude change in the frequency (i.e., from 1 MHz to ∼13

d[sine(ωt )]

the position of the state is minimal (i.e., |ωt =±90° ≈ 0), dt as in the green regions in Figure 2d, it is possible for electrons to tunnel resonantly between the source and the drain via the state and it is possible to observe a current. However, if this d[sin(ωt )]

variation in time is maximum (i.e., dt |ωt = 0° ≫ 1), as in the red regions, only negligible current can be observed. In Figure 3, we now turn to the impact of the rf on the response of the excited states of the donor atom. Figure 3a shows the excited state spectrum at the D+ to D0 transition with no rf signal applied, also consistent with previous measurements of this device.20,21,31 In this figure, the dc charge stability

Figure 3. Excited state spectrum at high rf frequencies. (a) Direct current gate stability diagram and (b) power dependence of the excited state spectrum when an rf excitation at ν = 12.785 GHz is applied to gate 1. (c) Same measurement as in (b) but with an rf excitation at ν = 10 GHz on gate 1 and both the gates addressed in dc as in Figure 2. The red ellipses and the red dashed lines in all sections of the figure outline the first excited state located at 10 ± 2 meV. 2446

DOI: 10.1021/acsnano.6b06362 ACS Nano 2017, 11, 2444−2451

Article

ACS Nano diagram focuses on the position of the first excited state, 1s(T2). Since we have the availability of two gates and the device is highly symmetric, the donor states can be capacitively addressed with both G1 and G2. As a consequence, we can address the states in two different regimes either when G1 and G2 are tied together and varied simultaneously or when G1 is fixed and G2 is varied. By addressing the donor in those two different regimes, we observe the same spectrum as in the original measurements20,21 where, as expected,33 the positions of each level are insensitive to the changes in electric field related to the different measurement configurations (see Figure 3a and Figure 5c). Obtaining the same results with different regimes of addressing the states is important as it demonstrates the reproducibility under different electric field conditions. In Figure 3b, we now present the same spectrum but with an rf excitation of ν = 12.785 GHz applied to G1. We observe the same doubling of current signature as observed in Figure 2c but now for the excited state spectrum, 1s(T2). This first excited state is located at 10 ± 2 meV, consistent with the previously measured bulk value for the first excited state of a single P donor in bulk silicon,20−22 as shown in Figure 1c. Likewise, in Figure 3c, we observe the same effect but now using an rf excitation at ν = 10 GHz applied to G1 and with both gates addressed in dc, as in Figure 2. These results are similar to those discussed in Figure 2d; however, this time the capacitive coupling is demonstrated for the 1s(T2) level of the excited state spectrum and shows robustness to the electric field across the donor. Quantum information and quantum metrology applications require precise and independent rf control of different gates,6,17,19 as many quantum logic gate operations include fast manipulation of electron states. These operations require the absolute synchronization in phase (time) between the rf signal individually applied to different gates and of their relative coupling to quantum states. To ascertain if this is possible in our device, in Figure 4, we present results obtained from

capacitive coupling between each gate and the donor state21 is in place and is preserved in the rf regime. These results show that, by precision STM patterning, it is possible to have control of the device symmetry and, as a result, to observe accurate nanosecond synchronization between different gates up to 0.25 GHz frequencies. The results presented so far are of relevance for the field of quantum computations as they demonstrate the control of energy states at f ≳ 10 GHz, that is, the high frequencies required for several quantum computer proposals which require synchronous sub-nanosecond pulses to be applied to quantum states.6,17 Precision transistors can also be used for single electron transfer applications, such as the ones necessary for quantum metrology,18,19 where independent and precise control in time of more than one gate is needed. In the next section, we shall show how, using excited state spectroscopy at ν = 50 MHz,23,24 we can distinguish the electron excited state spectrum of the donor from the 1D confinement-related DOS of the quasi-one-dimensional leads.20,21,31,32 As in the previous experiments, when we apply a square wave signal to one of the gates addressing the state, we observe a characteristic V shape of the current as a function of increasing pulse voltage (see Figure 5, where Vpulse represents the voltage amplitude provided to the bias tee). The V shape of the current represents the doubling of the ground state peak when square pulses are applied to G1 and is observed both for positive (Figure 5a) and for negative (Figure 5b) source bias voltages. This process is schematically described in Figure 5f for negative biases. In Figure 5a,b, the left branch shows the current where the ground state is pulsed from far above the bias window, while the right branch represents the dc ground state signature, which is shifted by the introduction of the pulse. There is an additional feature, labeled “#”, observed when a negative bias is applied to the source, as in Figure 5b, which we attribute to the first excited state of the donor electron as explained in the next section. It is worthwhile to remember that the DOS in the one-dimensional leads cannot be associated with this additional feature because the DOS signature is not Vpulse-dependent but only S/D bias-dependent.20,21,32 Hence, in these experiments, we can address both the excited and ground state spectrum at low bias such that pulse spectroscopy allows us to distinguish transport via the excited state and the DOS in the leads in a way not possible via dc spectroscopy.20,21 The Coulomb diamonds and the doubling observed in Figure 5a,b allow a direct conversion between gate voltage and energy. From the position of the red dot in Figure 5b at Vpulse = 120 ± 10 mV and using 0.075 for the final correction factor of the applied power (see the Methods section), we can determine an excited state energy of 9 ± 1 meV. This pulse-estimated value for the excited state energy 1s(T2) lies close to the one extracted from the dc data in Figure 5c, ≈10 ± 2 meV (black arrow and black dashed lines); see also red ellipses and red dashed lines in Figure 3b,c. The position of the other visible peak for the excited state (1s(E), white arrow and white dashed line around 13.5 meV) is also very close to the expected bulk values (i.e., 11.7 meV) and from previous estimations made of this device in the dc mode.20,21 It is important to understand why the overall dc excited state spectrum is more visible for negative bias (e.g., see Figure 5c), which indicates that the transparencies of the source/drain to the first excited state barriers (ΓSe/ΓDe) are asymmetric, with the latter being more transparent.21,34 The asymmetry in the tunnel barriers (i.e., ΓSe/ΓDe ≪ 1) can be better understood by looking at Figure 5g, where the negative bias regime is

Figure 4. Nanosecond synchronization between G1 and G2. A dc voltage is applied to G2, while a rf sine excitation of amplitude ≈ −8 dBm is provided to both G1 and G2. Vdc‑G1 and VSD are kept fixed at 600 and 0 mV, respectively. As expected, the maximum splitting of the peak is observed when the two rf signals are inphase (0° or 0 ns), while the minimum splitting is observed when they are out-of-phase (≈±180° or ±2 ns).

applying sinusoidal rf excitations of 250 MHz to the bias tees of both G1 and G2. Here, the provided rf excitations are of equal amplitude, but there is a varying difference in the absolute phase between the two signals. Hence, Figure 4 ultimately allows us to quantify the level of synchronization in time between the capacitive coupling between G1 and the GS and the one between G2 and the GS. The result of these measurements confirms that, within the limit of precision of the source (≈10 ps; see the Methods section), a very similar 2447

DOI: 10.1021/acsnano.6b06362 ACS Nano 2017, 11, 2444−2451

Article

ACS Nano

Figure 5. Estimation of the 1s(T2) excited relaxation rate. The observed V shapes represent the doubling of the ground state peak when square pulses with ν = 50 MHz, rise times of 90 ps, and a duty cycle of 50% are applied both for (a) >0 and for (b)